Genome-wide Analysis Identified SEMA4D, Novel Candidate Gene for Temperature Sensitivity in Patients With Non-Small Cell Lung Cancer

Abstract

Background:

In the era of precision medicine, individual temperature sensitivity has been highlighted. This trait has traditionally been used for cold-heat pattern identification to understand the inherent physical characteristics, which are influenced by genetic factors, of an individual. However, genome-wide association studies (GWASs) on this trait are limited.

Methods:

Using genotype data from 90 patients with advanced non-small cell lung cancer (NSCLC) and epidermal growth factor receptor mutations, we performed a GWAS to assess the association between single nucleotide polymorphisms (SNPs) and temperature sensitivity, such as cold and heat scores. The score of each participant was evaluated using self-administered questionnaires on common symptoms and a 15-item symptom-based cold-heat pattern identification questionnaire.

Results:

The GWAS was adjusted for confounding factors, including age and sex, and significant associations were identified for cold and heat scores: SNP rs145814326, located on the intron of SORCS2 at chromosome 4p16.1, had a P-value of 1.86 × 10⁻⁷; and SNP rs79297667, located upstream from SEMA4D at chromosome 9q22.2, had a P-value of 8.97 × 10⁻⁸. We also found that the genetic variant regulates the expression level of SEMA4D in the main tissues, including the lungs and white blood cells, in NSCLC.

Conclusions:

SEMA4D was found to be significantly associated with temperature sensitivity in patients with NSCLC, suggesting an increased expression of SEMA4D in patients with higher heat scores. The potential role of temperature sensitivity as a prognostic or predictive marker of immune response in NSCLC should be further studied.

Introduction

Pattern identification (PI), also known as syndrome differentiation or Zheng differentiation, is a fundamental concept in traditional East Asian medicine. ¹ It is a complex and individualized process based on the symptoms, signs, medical history, and other diagnostic information of a patient, used to identify a specific pattern of disharmony or imbalance in the body. ² Every individual is unique and, even if they have the same Western medical diagnosis, their PI can differ. ³ As a result, PI is a comprehensive medical condition that has traditionally played a role as a key code in diagnosis, treatment, and prognosis.^4,5 In recent years, there has been growing interest in integrative medicine, which seeks to combine elements of traditional practices, such as PI, with conventional Western medicine. This approach recognizes the potential benefits of both systems and aims to provide patients with a more comprehensive and personalized approach to their health and wellbeing.

In traditional Korean medicine (TKM), many types of PI are currently used, including “8 Principle PI,” “6 Meridian PI,” and “Visceral PI,” ⁶ of which 8 Principle PI is central in clinical practice due to its comprehensiveness. ⁷ Specifically, the 8 subtypes are yin, yang, exterior, interior, cold, heat, deficiency, and excess patterns. ⁸ Among the subtypes, the cold-heat PI, which is defined as temperature sensitivity, has been highlighted in the era of precision medicine. The heat pattern (HP) is characterized by dissatisfaction with hot temperatures, excessive sweating, elevated pulse rate, flushed face, restlessness, irritability, constipation, dehydration, and dark urine color. The cold pattern (CP) is characterized by a preference for warm temperatures, intolerance to cold, paleness, diarrhea, feeling cold in the extremities, and muscle tension. These patterns usually depend on a constitutional predisposition, which refers to the inherent physical and energetic characteristics of an individual influenced by genetic factors. A twin study also showed significant genetic effects on both HP and CP, where heritability estimates in both sexes were 40% for CP and 33% for HP, regardless of age. ⁹ Therefore, the biological causes of cold-heat patterns have attracted more scientific attention and have been widely studied as clinical biomarkers for various diseases. ¹⁰

Temperature sensitivity is commonly used for a personalized approach, particularly in cancer. ¹ For example, moxibustion in traditional treatments for cancer-related fatigue is more beneficial to patients with a CP than those with HP. ¹¹ Furthermore, temperature sensitivity might serve as a prognostic or predictive marker in patients with cancer. In detail, gefitinib, a tyrosine kinase inhibitor (TKI), was more effective for patients with non-small cell lung cancer (NSCLC) and CP than those with non-CP. ¹² Interestingly, immune checkpoint inhibitors (ICIs) significantly prolonged the progression-free survival (PFS) in patients with NSCLC and non-CP than those with CP. ¹³ Moreover, patients with NSCLC and CP rather than HP were more likely to have epidermal growth factor receptor (EGFR) gene mutations. ¹⁴

With the availability of human genomic data, links between temperature sensitivity and candidate genes have been reported. Two subtypes in patients with rheumatoid arthritis showed significantly different gene expression and metabolite profiles for the regulation of apoptosis. ¹⁵ Recently, a genome-wide association study (GWAS) of a population-based cohort suggested that the CIRP gene is associated with CP. ¹⁶ However, reports on genetic analysis with temperature sensitivity are still limited. Moreover, a GWAS in diverse cohorts is necessary for the discovery of candidate genes that lead to the disclosure of the genetic implications of temperature sensitivity. Here, we explored genetic associations for temperature sensitivity using the symptom-based cold-heat PI questionnaire (CHPIQ) in patients with advanced NSCLC.

Materials and Methods

Study Design

Ninety patients were enrolled in the HangAmDan-B1-Investigator Initiated Trial (HAD-B1-IIT) study, which was a randomized, multicenter, open-label study for the comparison of the safety and efficacy of afatinib monotherapy and that of the combination therapy of afatinib and HAD-B1 for patients with locally advanced NSCLC and EGFR mutations. The participants were recruited between February 2021 and April 2023. The Institutional Review Board of Kosin University Gospel Hospital reviewed and approved this study (IRB No. KUGH2020-05-037). All patients understood the study, voluntarily agreed to participate, and signed a written consent form for the clinical trial. They were enrolled in our study across 5 hospitals: Kosin University Gospel Hospital (N = 31), Pusan National University Yangsan Hospital (N = 16), Konyang University Hospital (N = 8), Seoul St. Mary’s Hospital (N = 30), and Ajou University Hospital (N = 5). The temperature sensitivity of the participants was evaluated using self-administered questionnaires on common symptoms and a 15-item (8 questions for CP and 7 for HP) symptom-based CHPIQ. ¹⁰ On a Likert scale of 1 to 5, each symptom was assessed. The score ranged from 8 to 40 points for CP and 7–points for HP, with a higher score indicating either CP or HP (Supplemental Table 1).

Genotyping, Imputation, and Quality Control

To extract DNA from 90 patient blood samples (41 males and 49 females), THERAGEN BIO (Seongnam, Korea) used 260/280 nm and 260/230 nm ratios to determine DNA quality and quantity. DNA samples were stored at 80°C until further processing. All DNA samples were amplified and randomly fragmented into 25 to 125 bp fragments, purified, resuspended, and hybridized to an Axiom array (TPMRA chip, Thermo Fisher Scientific, Waltham, MA, USA) that was developed based on the Asian Precision Medicine Research Array (Thermo Fisher Scientific). After hybridization, unbound DNA fragments were washed away to reduce nonspecific ligation noise. For each patient sample, 820 000 single nucleotide polymorphisms (SNPs) were genotyped for a genome-wide coverage.

After quality control (QC) of the genotyping results according to the manufacturer’s instructions, we conducted imputations using the Minimac program with a reference panel of East Asian samples (n = 536) from the 1000 Genome Project Phase III integrated variant dataset.^17,18 The 1000 Genomes data used in the imputation procedure can be download from the following FTP link: ftp://share.sph.umich.edu/minimac3/G1K_P3_M3VCF_FILES_WITH_ESTIMATES.tar.gz.

We based our gene annotation and SNP locations on the NCBI Human Genome Build 38 (hg38). Variations and samples with call rates of less than 95%, a Hardy-Weinberg equilibrium failure of less than 1 × 10⁻⁶, and minor allele frequencies of less than 1% were removed. Finally, 1 723 163 SNPs in 89 patients passed the filters and QC and were retained and carried forward for GWAS analyses.

Statistical Analysis

Demographic characteristics and clinical features between the groups were compared using t-tests or chi-square tests, as appropriate, using R statistical software (version 4.03). We performed a genome-wide association analysis of the scores for cold or heat sensitivity with adjustments for sex, age, and institutes using PLINK version 1.90 (https://www.cog-genomics.org/plink/, accessed on September 6, 2022). Due to the small sample size, logistic regression for CP or HP was not performed. In addition, significant SNPs were chosen using the P-value threshold (<1 × 10⁻⁷), borderline genome-wide significance. ¹⁹ Expression quantitative trait loci (eQTL) were explored for the identified SNPs in the gene expression data based on GTEx release version 8. For this set of analyses, 2-tailed P-values < .05 were considered statistically significant.

Results

Descriptive Characteristics of NSCLC Patients

The characteristics of the 89 study participants are summarized in Table 1. The median age of 40 males and 49 females was 69 years. Based on the symptom-based CHPIQ scores, 31 (35%) patients had HP and 58 (65%) patients had CP. As expected, the score for cold sensitivity was significantly higher CP group whereas the score for heat sensitivity was significantly higher HP group (P-value < .0001). Mean ages were higher in CP group (69.9 ± 10.8) than in HP group (67.0 ± 9.0) but it was not statistically significant (P-value = .187). There were no differences between the CP and HP groups with respect to sex, BMI, performance status, stage, metastatic sites, clinical features, or laboratory parameters (all P-value > .05). The prevalence of hyperlipidemia was significantly higher in the HP group (P-value = .0173). The acute inflammatory protein C-reactive protein (CRP) ²⁰ level was slightly higher in the HP group, although the difference was not statistically significant (P-value = .204).

Table 1.

Demographic Characteristics and Clinical Features Between Cold and Heat Pattern.

	Total participants (n = 89)	Cold pattern (n = 58)	Heat pattern (n = 31)	P-value*
Female, n (%)	49 (55.1)	33 (56.9)	16 (51.6)	.800
Age (years)	68.9 ± 10.3	69.9 ± 10.8	67.0 ± 9.0	.187
BMI (kg/m2)	23.7 ± 3.1	23.7 ± 3.1	23.8 ± 3.1	.813
Cold_Score	22.2 ± 5.9	24.4 ± 5.7	18.2 ± 3.7	<.0001
Heat_Score	16.6 ± 5.3	14.5 ± 4.4	20.5 ± 4.5	<.0001
Performance Status by ECOG, n (%)
1	17 (19.1)	9 (15.5)	8 (25.8)	.187
2	63 (70.8)	41 (70.7)	22 (71.0)
3	9 (10.1)	8 (13.8)	1 (3.2)
Stage, n (%)
Locally advanced	6 (6.7)	4 (6.9)	2 (6.4)
Distant metastases	83 (93.3)	54 (93.1)	29 (93.6)	.936
Distant metastatic sites, n (%)
0	6 (6.8)	4 (6.9)	2 (6.5)
1	39 (43.8)	28 (48.3)	11 (35.5)
2	44 (49.4)	26 (44.8)	18 (58.0)	.478
Comorbidity, n (%)
Hypertension	41 (46.1)	27 (46.6)	14 (45.2)	.900
DM	19 (21.3)	14 (24.1)	5 (16.1)	.380
Hyperlipidemia	19 (21.3)	8 (13.8)	11 (35.5)	.0173
Treatment history, n (%)
Surgery	20 (22.5)	10 (17.2)	10 (32.3)	.106
Chemotherapy	10 (11.2)	7 (12.1)	3 (9.7)	.734
Radiation	5 (5.6)	3 (5.2)	2 (6.5)	.803
Hematological test
WBC (103/µL)	8.17 ± 2.79	7.91 ± 2.53	8.67 ± 3.17	.260
Hgb (g/dL)	13.01 ± 1.48	12.87 ± 1.51	13.28 ± 1.39	.212
PLT (104/µL)	293.9 ± 89.4	291.1 ± 89.1	299.1 ± 89.7	.691
ESR (mm/h)	25.0 ± 22.4	24.2 ± 22.3	26.5 ± 22.4	.637
Chemistry test
AST (U/L)	24.6 ± 14.0	25.0 ± 15.9	23.9 ± 9.5	.684
ALT (U/L)	25.4 ± 41.1	26.6 ± 50.0	23.2 ± 13.2	.627
BUN (mg/dL)	15.29 ± 6.23	15.15 ± 6.13	15.56 ± 6.40	.770
Cr (mg/dL)	0.75 ± 0.19	0.75 ± 0.21	0.76 ± 0.17	.800
CRP (mg/dL)	1.43 ± 2.39	1.17 ± 2.06	1.93 ± 2.85	.204
CEA (ng/mL)	291.1 ± 668.4	268.2 ± 611.1	333.9 ± 762.2	.685

Data are presented as n (%) or mean ± standard deviation.

Abbreviations: BMI, body mass index; ECOG, Eastern Cooperative Oncology Group; DM, diabetes mellitus; WBC, white blood cell; Hgb, hemoglobin; ESR, erythrocyte sedimentation rate; PLT, platelet; AST, aspartate aminotransferase; ALT, alanine aminotransferase; BUN, blood urea nitrogen; Cr, creatinine; CRR, C-reactive protein; CEA, carcinoembryonic antigen.

Genome-Wide Analyses for Cold or Heat Score as Temperature Sensitivity

After filtering SNPs and a sample as described in the Methods section, genome-wide outcomes for cold or heat scores are shown in quantile–quantile and Manhattan plots (Figure 1). The cold score showed no SNPs over the borderline genome-wide significance, whereas the SNP rs145814326 with a P-value of 1.86 × 10⁻⁷ was the most significant and was located on the intron of SORCS2 on chromosome 4p16.1 (Figure 1E). In comparison, the heat score had the highest association with rs79297667 located upstream from SEMA4D at chromosome 9q22.2, with a P-value of 8.97 × 10⁻⁸ (Figure 1F). The best associations are summarized in Supplemental Table 2.

Figure 1.

The distribution of cold or heat score based on each questionnaire in 89 patients (A and B). Genome-wide analyses according to cold or heat score were shown in Quantile-Quantile plot (C and D) and Manhattan plot (E and F). The red line indicates a borderline genome-wide significant P-value of 1 × 10⁻⁷.

The eQTL Analysis for the Candidate Genes: SORCS2 and SEMA4D

Because significant loci, such as rs145814326 and rs79297667, were not available in the GTEx database, we explored proxy variants in the Japanese population of 1000 genomes in East Asia, which are most similar to the Korean population. Proxy variants with top associations in linkage disequilibrium (LD) are summarized in Supplemental Tables 3 and 4. No significant eQTLs were found for the LD variants of rs145814326 in any tissue. However, a strong eQTL relationship for the variants in LD with rs79297667 was observed with the expression levels of SEMA4D (the top 5 SNPs are shown in Figure 2). For example, the minor allele of rs62547238 leads to a significant increase in SEMA4D levels in whole blood (P-value = 3.16 × 10⁻²⁰, normalized effect size = 0.13), spleen (P-value = 5.18 × 10⁻⁸, normalized effect size = 0.26), and lung (P-value = 2.33 × 10⁻⁶, normalized effect size = 0.14). Taken together, patients with higher heat scores were likely to express SEMA4D in tissues related to lung and white blood cells due to the minor allele of rs79297667.

Figure 2.

The significant eQTLs for the top 5 proxy variants of rs79297667 shown in violin plots; rs12686034 (A), rs4877088 (B), rs62547240 (C), rs56177051 (D), and rs62547238 (E). The minor allele in each variant significantly upregulates the level of SEMA4D in the lungs, skeletal muscles, spleen, and whole blood.

Discussion

To the best of our knowledge, this is the first GWAS on temperature sensitivity in patients with NSCLC. In this analysis, we found that the minor allele of rs79297667 located upstream of SEMA4D was significantly associated with heat sensitivity score. We also suggested that a higher heat score induces the expression of SEMA4D as a novel gene for temperature sensitivity, based on a significant eQTL for the SEMA4D gene. Interestingly, the tissues observed in the eQTL association were the lungs and white blood cells, which are closely related to pathological implications in NSCLC.

Semaphorin 4D (SEMA4D), also called Cluster of Differentiation 100 (CD100), is a member of a large semaphorin family that includes signaling molecules. ²¹ Semaphorins are classified into two main classes based on their structural characteristics and signaling mechanisms. Class 3 semaphorins are secreted proteins that primarily act as axon guidance cues during neural development, ²² whereas class 4 semaphorins are transmembrane proteins that play roles in both axon guidance, immune regulation, angiogenesis, and tumor progression.^23,24 Soluble SEMA4D (sSEMA4D) is a truncated form of the transmembrane semaphorin 4D protein, which is generated via proteolytic cleavage of the full-length transmembrane SEMA4D.^25,26 Unlike the full-length transmembrane form, sSEMA4D lacks the transmembrane and cytoplasmic domains, making it freely soluble in the extracellular environment. In recent years, research has focused on understanding the immune-regulatory functions of sSEMA4D in the context of inflammatory diseases and cancer.^{23 -25} sSEMA4D has been identified as a key immune-regulatory molecule involved in modulating immune responses. ²⁵ Therefore, it affects the behavior of various immune cells, including T cells, B cells, dendritic cells, and macrophages, where it either promotes or inhibits immune cell activation and function depending on the specific context and microenvironment. ²⁷ In the pathogenesis of various inflammatory diseases, such as rheumatoid arthritis, multiple sclerosis, and inflammatory bowel disease, sSEMA4D appears to contribute to immune cell activation and cytokine production, thereby leading to increased inflammation. ²⁸ In the context of cancer immunology,^23,29 s SEMA4D influences the tumor microenvironment, affects immune cell infiltration and activity, and impacts the anti-tumor immune response, which promotes tumor growth and metastasis by modulating the immune response in favor of the tumor. ³⁰ Furthermore, it is involved in the progression of cancers including breast cancer, colorectal cancer, and lung cancer. ³¹

From the viewpoint of TKM, HP is likelier than CP to generate excess heat from the body via cell metabolism, where temperature sensitivity may be associated with the metabolic rate of the body. A clinical study of healthy Koreans also showed that participants with higher CP scores had a lower resting metabolic rate than those in the same age-sex group with lower scores. ³² Likewise, it could be inferred that the immune system is more active in HP. With the advent of immune checkpoint inhibitors (ICIs), “cold tumor” and “hot tumor” are terms used in the context of cancer immunology to describe the level of immune activity within a tumor and its microenvironment. ³³ Hot tumors are generally associated with elevated immune activity and are more responsive to ICIs, whereas cold tumors have limited immune responses and may be less responsive to these treatments. ³⁴ Previously, patients with NSCLC and CPs were reported to be poorly responsive to ICIs because they had a relatively small number of active CD8 T cells based on immune profiling assay in the blood. ¹³ In the patients with NSCLC herein, CRP was slightly higher in HP group, and our genomic analysis suggested that a higher heat score increases the expression of SEMA4D for cancer progression. Taken together, temperature sensitivity indicates the immune status for cancer progression and may be used as a prognostic or predictive marker for ICIs in NSCLC.

Our study had several limitations. First, the sample size was small and replication was difficult to implement. The statistical power was poor compared with that of conventional GWASs. We selected borderline genome-wide significance (<1 × 10⁻⁷) as the P-value threshold, instead of standard significance (<5 × 10⁻⁸). We could not replicate our findings in another cohort because the temperature sensitivity trait had not yet been widely collected. Second, temperature sensitivity may be influenced by various confounding factors, such as age, sex, and cancer treatment. For example, in general, aged participants and females are likely to have CP rather than HP. ³² Therefore, we adjusted for age and sex as covariates in the GWAS even though all confounding factors could not be included. Third, the association between SEMA4D expression and prognostic variables in NSCLC has not been tested. Finally, an objective evaluation of cold-heat PI is challenging. The diagnosis of the cold-heat PI was largely based on the subjective symptoms of an individual and the intuitive knowledge of the doctor ³⁵ ; thus, a reliable and valid questionnaire is essential for research regarding temperature sensitivity. ¹⁰ For this reason, the CHPIQ had been developed in several forms depending on the purpose, ranging from questionnaires based on diseases to that for usual symptoms. ¹⁰ We used a 15-item symptom-based CHPIQ that had been officially approved as the standard method for the temperature sensitivity in TKM.^10,36

In conclusion, we found that SEMA4D was significantly associated with temperature sensitivity in patients with NSCLC, suggesting an increased expression of SEMA4D in patients with higher heat scores. The potential role of SEMA4D as a clinical biomarker of immune function should be further explored, leading to the implication of temperature sensitivity on the immune status of NSCLC.